Analysis of the N-glycosylation profiles of the spike proteins from the Alpha, Beta, Gamma, and Delta variants of SARS-CoV-2

COVID-19, caused by SARS-CoV-2, continues to evolve with numerous spike (S) mutations that can enhance infectivity and immune evasion. The trimeric spike mediates ACE2 binding and membrane fusion and is heavily glycosylated at 22 N-glycosylation sequons per protomer. Glycans contribute to folding, stability, conformational dynamics (open/closed RBD), and immune shielding. Prior studies established predominant complex-type N-glycans with oligomannose at select sites (e.g., N234). This study asks whether VOC-associated spike mutations (Alpha, Beta, Gamma, Delta) alter N-glycosylation profiles compared to D614G, potentially impacting antigenicity, receptor engagement, and vaccine design.

Previous mass spectrometry studies of recombinant and virion-derived SARS-CoV-2 spike demonstrated occupancy at all 22 N-sequons, with predominant complex-type glycans and oligomannose enrichment at sites such as N234 and, often, N61. Molecular dynamics and mutagenesis indicated specific glycans (e.g., N165, N234, N331/N343) modulate RBD conformation and viral entry, and perturbing N-glycosylation reduces infectivity. Reports comparing variants generally found conserved glycan processing with site-specific differences; discrepancies across laboratories were noted and attributed to protein source, analytical conditions, and data processing. These foundations motivated a standardized, multi-digestion LC-MS/MS approach to compare variant spikes produced under uniform conditions.

Recombinant His-tagged, trimeric 2P-stabilized ectodomains (mutated furin site; K986P, V987P) of S variants (D614G, Alpha, Beta, Gamma, Delta) expressed in HEK293 cells were sourced from one manufacturer (R&D Systems). Proteases (trypsin, chymotrypsin, Lys-C, Asp-N, alpha-lytic protease) were used in single and sequential digestions to generate single-sequon glycopeptides. Typically, 1.5–2 µg protein was denatured/reduced (60 °C; 50 mM ammonium bicarbonate, 0.05% RapiGest, 5 mM DTT), alkylated (15 mM iodoacetamide, 30 min, dark), then digested overnight at 37 °C (1:10 enzyme:protein, w/w). Sequential digests used a first 1 h step at 52 °C (trypsin, Lys-C, or Asp-N) followed by overnight digestion at 37 °C with a second enzyme (e.g., chymotrypsin, alpha-lytic protease, Asp-N). Reactions were quenched with formic acid (pH<3), RapiGest precipitated, centrifuged, and analyzed in technical triplicate using three protein batches from identical preparations. LC-MS/MS used an Orbitrap Eclipse Tribrid with UltiMate 3000 RSLCnano. Peptides were trapped (PepMap 100, 75 µm × 2 cm) and separated on an EASY-Spray C18 column (75 µm × 15 cm, 2 µm) with a multi-step gradient at 300 nL/min. MS1 scans: 120,000 resolution (m/z 200), m/z 375–2000. Data-dependent HCD MS2 (28% NCE, Orbitrap 30,000 resolution) with a signature oxonium ion trigger (HexNAc 204.0867, 138.0545, 366.1396 within 15 ppm) initiated follow-up EThcD MS2 (charge-dependent ETD with 35% supplemental HCD; Orbitrap 50,000 resolution; AGC and injection times as specified). Data analysis used Byonic (PMI 3.7) in Proteome Discoverer 2.4 with semi-specific cleavage, up to 3 missed cleavages, precursor tolerance 6 ppm, fragment tolerance 20 ppm (HCD/EThcD). Fixed carbamidomethyl (C); variable deamidation (N/Q) and oxidation (M) were allowed. Searches used the PMI human N-glycan (182, no multiple fucose) database with up to two N-glycans per peptide. Protein FDR 1% or 20 reverse count; glycopeptide identifications filtered at Byonic score ≥150. Relative abundance per glycan type at each site was calculated as normalized peak intensity of that glycoform over total glycopeptides for that site, averaged across three replicates (SD <30%). To ensure robust quantification, datasets per site were selected based on coverage criteria across digestion conditions (≥80% total intensity for dominant peptide sequence(s)), favoring conditions with more glycoforms and confirmed trends across at least two digestion methods.

- Glycopeptides covering all 22 conventional N-glycosylation sites were detected in general; however, quantification was not possible for some: N17 is absent in Delta (T19R mutation removes the sequon), N17 in Gamma was unresolved due to co-occurrence with N20, and N717 in Delta and N149 across variants had insufficient spectral quality for quantitation. - Gamma harbors two novel sequons, N20 (T20N) and N188 (R190S), both observed to be occupied. N20 was predominantly complex-type (~80% complex; high fucosylation ~92% and sialylation ~29%), while N188 was mainly oligomannose (~71%) with low fucosylation/sialylation. - Across 21 evaluable conventional sites, complex-type glycans predominated (>80% occupancy) at most sites for all variants, except for N61, N234, and sometimes N717. Oligomannose remained dominant at N234 across all variants and was also prevalent at N61. - Site-specific complex glycan ranges (Table 3) showed high similarity across variants but with notable differences at selected sites. For example, Delta exhibited reduced complex occupancy at several sites, notably N122 (Delta ~58.7–59% complex vs 84–93% in others). Many sites displayed limited inter-variant variation (e.g., N74, N282, N331, N657, N709, N1098, N1134, N1158, N1173), whereas others (N122, N165, N603, N616, N717) showed more divergence in glycoform distributions. - Dominant glycoforms at multiple sites included monofucosylated biantennary complex structures with or without sialic acid (N4H5F1, N4H5F1A1). N61 was dominated by N2H5, and N234 by N2H5–N2H8, consistent with oligomannose enrichment. - Microheterogeneity: the number of distinct glycans per site ranged from ~24 to 91. Sites N61, N74, N234, N717, and N1173 showed lower heterogeneity (<40 glycoforms), whereas N17, N282, N1074, N1098, and N1194 exhibited higher diversity (~80 glycoforms). - Fucosylation and sialylation trends varied by site: N61 and N234 had particularly low sialylation/fucosylation (~5% and ~20%, respectively), N1098 was highly sialylated (70.5–77.1%) with lower fucosylation (22.4–36.0%), while N343 and N603 were more fucosylated (68.4–79.6% and 70.5–92.8%) than sialylated (14.0–21.9% and 14.7–34.2%). - Methodological insight: no single protease condition sufficed for all sites; Lys-C, alone or in combinations, yielded broadest coverage. Selection of peptide sequence(s) per site (often two closely related sequences) was essential for accurate quantitation and helps explain inter-lab discrepancies. Overall, despite amino acid changes across Alpha, Beta, Gamma, and Delta, site-specific N-glycan processing states and compositions were largely conserved relative to D614G, with targeted differences at select NTD and S2 sites.

The study addressed whether spike mutations in VOCs alter N-glycosylation profiles sufficiently to impact structure, receptor binding, or immune shielding. The results show that while certain sites (e.g., N122, N165, N717, N1158, N1173) exhibit variant-dependent shifts in glycoform distributions and processing, the global pattern of spike glycosylation remains highly conserved from D614G through Delta: most sites are complex-type, with consistent oligomannose at N234 (and N61). Such conservation implies that the overall glycan shield architecture and its functional roles in stability and immune evasion are maintained across these variants. Differences at functionally important sites, such as N165 (involved in RBD conformational control) and N122/NTD sites (implicated in neutralizing antibody interactions), may contribute to altered spike dynamics, antigenicity, or ACE2 engagement, providing plausible mechanistic underpinnings for variant phenotypes. Detection of two novel Gamma sequons (N20 and N188) and their distinct processing (complex vs oligomannose) highlights how mutations can add or remove glycan sites, potentially affecting epitope shielding and spike stability. The methodological framework—uniform protein source, multi-protease digestion, and cross-confirmation across digests—strengthens confidence in site-specific quantitation and clarifies sources of discrepancy in the literature.

Using advanced LC-MS/MS with multi-enzyme digestion and stringent data selection, the study mapped and quantified N-glycans at most sequons of SARS-CoV-2 spike from Alpha, Beta, Gamma, Delta, and D614G. Mutations influenced glycan abundance and types at select sites (notably in the NTD and S2), and Gamma introduced two novel occupied sequons (N20, N188). However, the overall glycosylation landscape—glycan types, occupancy, heterogeneity, and fucosylation/sialylation trends—was highly similar across variants, indicating conserved glycosylation crucial for spike structure/function and immune shielding. These findings suggest that, despite sequence evolution, the conserved glycan shield may support continued vaccine effectiveness, though subtle compositional differences at key sites could influence antigenicity and warrant further study. Future work should extend to complete coverage of challenging sites, detailed structural glycan features (linkage, LacdiNAc, M6P), O-glycosylation, and assessments across protein lots and acquisition/analysis pipelines.

Some glycosylation sites (e.g., N149, N17 in Gamma, N717 in Delta) could not be robustly quantified due to low spectral quality, overlapping sequons, or missing sequons. Reliance on selected datasets per site (≥80% intensity coverage) across digestion conditions may omit low-abundance peptides. Structural glycan features (linkages, LacdiNAc, M6P) were not resolved by the applied methods, potentially biasing interpretation of complex-type abundance in HEK-expressed proteins. Results may be affected by protein source/lot, MS acquisition parameters, and data processing choices; even within one manufacturer, lot-to-lot variation can occur. Additional proteases/conditions and targeted methods are needed to improve coverage, characterize O-glycosylation and other modifications, and refine structural glycan assignments.